Most Expert Advisors that fail in live trading compiled without errors and backtested profitably. The problems are structural — silent order failures, missing state persistence, hardcoded assumptions that only work in the Strategy Tester — and they only surface when real money is on the line. At barmenteros FX, we audit your EA’s source code, identify every structural risk, and either fix it or tell you honestly whether a rebuild makes more sense.
What You Get
A complete forensic review of your EA’s MQL4 or MQL5 source code, delivered as a structured report with specific, actionable findings:
- Error handling audit — every `OrderSend()` / `trade.PositionOpen()` call checked for return value validation, retry logic, and error logging
- State persistence review — does your EA survive terminal restarts, VPS reboots, and broker disconnects without losing track of open positions, grid levels, or trade counters
- Execution logic verification — entry/exit signals tested against broker-specific constraints: stop levels, spread widening during news, trade context busy conditions
- Risk management validation — lot sizing, maximum drawdown guards, margin checks, and position limits verified against real-world edge cases
- Strategy logic assessment — backtest-to-live discrepancy analysis, overfitting indicators, and out-of-sample viability
- Prioritized fix list — every issue ranked by severity (critical / important / minor) with specific code locations and recommended solutions
You receive the full report regardless of whether we do the fixes. The report is yours — use it to fix the code yourself, hand it to another developer, or commission us to implement the changes.
How It Works
Step 1 — Submit your EA and describe the problem. Send your `.mq4` or `.mq5` source file through our quote form. Tell us what the EA is supposed to do, what is going wrong (or what you suspect might go wrong), and whether it has been running live.
Step 2 — We review the code. We read every line. Not a surface scan — a structural analysis of error handling, state management, execution logic, and risk controls. We test against the five failure patterns that appear in the majority of rescue projects: silent `OrderSend`, amnesia on restart, hardcoded assumptions, ignored execution feedback, and missing defensive guards.
Step 3 — You receive the report. A detailed document with findings, severity ratings, and specific recommendations. For each issue, we explain what can break, under what market conditions, and what the fix looks like.
Step 4 — Decide what happens next. Fix it yourself, hire someone else, or commission us. If we do the fixes, the review cost is credited toward the repair quote.
Who This Is For
This service exists for traders and firms in specific situations:
- Your EA backtests profitably but behaves differently live — requotes, missed entries, unexpected position sizes, orders that should have closed but did not
- You bought or commissioned an EA and want an independent second opinion before risking real capital
- AI generated your EA code (ChatGPT, Copilot, Claude) and you need a human who understands broker execution to verify it will survive live markets
- Your EA crashed or lost money after a terminal restart and you need to understand why
- You are scaling from demo to live and want to know every structural risk before real money is on the line
Who This Is Not For
- Traders looking for someone to create a strategy — we review and fix code, not develop trading ideas. For full EA development, see our custom EA development service.
- EAs without source code — we need the `.mq4` or `.mq5` file. We cannot review compiled `.ex4` / `.ex5` files.
What barmenteros FX Finds in Code Reviews
After 14 years of building and rescuing trading systems, I see the same structural failures in roughly 80% of the EAs that arrive at my desk. These are not obscure bugs — they are predictable, systematic gaps that the Strategy Tester never reveals.
Silent order failures. The EA calls `OrderSend()` and never checks the return value. In backtest, every order fills instantly. In live trading, requotes happen during fast markets, margin runs out after drawdowns, and broker rejections occur during restricted sessions. The EA does not know the order failed, updates its internal state based on a position that does not exist, and compounds errors from that point forward.
Terminal restart amnesia. Every runtime variable resets when MetaTrader restarts. Grid levels reset to zero. Daily trade counters reset. Trailing stop positions are forgotten. A martingale EA mid-recovery at 8x lot size restarts at 1x, opening a base position instead of the recovery position — doubling exposure without the recovery logic that was supposed to manage it.
Hardcoded broker assumptions. Pip values, stop levels, lot step sizes, and symbol naming conventions hardcoded for one specific broker. Move the EA to a different account — or the broker updates its specifications — and calculations silently produce wrong values.
Ignored execution feedback. The EA sends a modification command and does not verify it was applied. `OrderModify()` returns `false`, the stop loss was not actually moved, but the EA proceeds as if it was. The position runs unprotected.
At barmenteros FX, we see these patterns in roughly 4 out of 5 rescue projects over the past three years — across 30+ code reviews.
Pricing and Next Steps
Code review pricing depends on the EA’s complexity:
| Complexity | Typical Scope | Price Range |
|---|---|---|
| Simple | Single-strategy EA, <500 lines, basic order management | $150 – $250 |
| Medium | Multi-condition EA, grid/martingale logic, 500–1,500 lines | $250 – $450 |
| Complex | Multi-pair, multi-timeframe, external integrations, 1,500+ lines | $450 – $800 |
Fixes and repairs are quoted separately after the review, based on findings. The review cost is credited toward repair work.
To get started: Submit your EA for review →
Include your source file and a brief description of what the EA does and what concerns you have. We will confirm the complexity tier and provide a fixed quote before any work begins.
Frequently Asked Questions
Do I need to provide the source code?
Yes. We need the `.mq4` or `.mq5` source file. Compiled `.ex4` or `.ex5` files cannot be reviewed — they contain machine code, not readable source. If you do not have the source, the original developer should provide it. At barmenteros FX, we include full source code delivery on every project we build.
Can you review AI-generated EA code?
Yes — and this is increasingly common. AI tools like ChatGPT and GitHub Copilot can generate MQL code that compiles and backtests, but they consistently miss broker execution constraints, state persistence across restarts, and defensive guards that only matter in live trading. barmenteros FX specializes in making AI-generated trading tools actually work in production. We review the generated code against the same standards we apply to human-written EAs.
What if the EA needs a complete rebuild?
We will tell you directly. The review report includes a fix-or-rebuild recommendation with a clear justification. If the code structure cannot absorb targeted fixes without cascading regressions, rebuilding from the working strategy logic is often faster and cheaper than patching. We explain exactly why in the report, so you can make an informed decision.
How long does a code review take?
Typically 3–5 business days from receiving the source code. Complex EAs with multiple interconnected systems may take up to 7 business days. We confirm the timeline before starting.
Is my trading strategy kept confidential?
Absolutely. All code reviews are covered under our standard confidentiality terms. We do not share, publish, or reuse client code or strategy logic. Your intellectual property remains yours.
Related: Custom EA Development · Our Services · Request a Quote